FastClaw / Run the model

model supplier

FastClaw supports OpenAI, Anthropic, Ollama, OpenRouter, Groq, DeepSeek, Mistral, and OpenAI-compatible APIs.

Supported Providers

Both the README and source code design providers as replaceable layers. Built-in presets cover mainstream LLM services and support any OpenAI-compatible endpoint.

OpenAI / Anthropic

Ideal for production-quality conversations, tool calls, and multimodal input.

Ollama

Suitable for local models and offline experiments.

OpenRouter / Groq

Suitable for multi-model routing, high-speed inference or low-cost invocation.

DeepSeek / Mistral

Suitable for specified model families and regionalized supplier access.

Model naming

Agent Manager will parse model into <providerKey>/<modelId>. For example, openai/gpt-4o-mini uses the openai provider and the model ID is gpt-4o-mini.

fastclaw agents config alpha set model openai/gpt-4o-mini
fastclaw agents config alpha set model anthropic/claude-sonnet-4
fastclaw agents config alpha set model ollama/qwen2.5-coder

Credential source

providerForAgent first reads the per-agent provider credential, and then fallsback shared provider. The plug-in provider will also get per-call credentials, and the source code requires the plug-in not to cache the credentials.

Recommended practices

When connecting to multi-tenants or upstream SaaS, do not share the same provider key with all Agents. Partitioning the provider config by user, agent, or line of business makes auditing and quotas easier.